I have a vendor supplied Java console application for loading historical data into their webapp and an Oracle 11g backend all runing on redhat. Performance testing is projecting upwards of a month or more to load our volume of historical data.
I have no access to the source code, and can only configure the connection in property files. The documentation specifies using the oracle thin client - I can also get it to run with the OCI client but that is slower still. I've looked at ADDM reports on the DB and consistently it recommends using connection pooling as its number one finding.
Is there a way to fake out or simulate connection pooling - make the application think its requesting a connection from the DB but its coming from a custom pooling layer?
I've also tried increasing the heap on the java app but it doesn't even use a fraction of what I give it, and doesn't seem to be multithreaded at all.
Any tips or suggestions are welcome.